Viktor Malík

> do we really want `_` to be a token or should we use `*` instead? I'd go with `*`, that reflects better the purpose of the syntax (matching any...

> IMHO, these ids are used to reference format arguments in the data maps, and there are different kind of maps for arguments. It is like we have a one-to-one...

Ok then, I'll look into this and try to integrate DWARF support. What about using libdw (from elfutils) instead of libdwarf? It's supposed to be faster, has a nicer API,...

> I haven't seen it mentioned anywhere so I'll do it here; any thoughts on adding something like #1966 for USDT probes? Simply listing the number of arguments and possibly...

> That leaves the problem of having two interfaces that basically do the same thing. Imagine telling a user "on newer kernels change all your kprobes to kfunc and it'll...

I see. Yeah, this could make more sense in the future since at this moment, we could use kfunc behind the scenes only for probes attached to a single function...

Checked, all seems to be working as expected. The only thing that I found is that `tools/bitesize.bt` is broken on my kernel. I have a fix ready (#2236), not sure...

I can see the use-case for this (portability of bpftrace scripts) but I wonder if we can come with some more compact syntax (which I believe also @fbs meant by...

Unfortunately, Ubuntu 20.04 doesn't package libbpf, which is now required for bpftrace. The solution is to build libbpf from source, as is done in our [Docker build script](https://github.com/iovisor/bpftrace/blob/master/docker/build.sh#L28-L35).

> Where should i add more tests. Some unit tests in `tests/semantic_analyser.cpp` would be useful. Something similar to what we have for other calls.